09:35 <DavidHeidelberg> is Jonas Schwöbel around?

09:58 <jenneron[m]> David Heidelberg: ask, i have his contact

09:58 <DavidHeidelberg> just missing his driver Documentation ;)

09:58 <jenneron[m]> right, it may not exist

09:59 <jenneron[m]> i sent him your message

10:06 <DavidHeidelberg> thx jenneron

10:08 <DavidHeidelberg> anyone bored? these txt's waiting for yaml conversion [ad,tegra-audio-plutux, nvidia,tegra-audio-wm8903] [ad,tegra-audio-wm8903-medcom-wide, nvidia,tegra-audio-wm8903] [ad,tegra-audio-wm8903-tec, nvidia,tegra-audio-wm8903] [nvidia,tegra114-ahb, nvidia,tegra30-ahb] [nvidia,tegra114-ahub] [nvidia,tegra114-apbdma] [nvidia,... (full message at https://libera.ems.host/_matrix/media/r0/download/libera.chat/af8d4bcf734a8ed8d488a74a8ae2b6d90433947b)

12:03 <pgwipeout[m]> Hey, quick question, does anyone know why u-boot on T30 will work fine unencrypted, but fails to execute when encrypted? Is there a special header I need to add to it or something for it to load correctly when encrypted?

12:03 <pgwipeout[m]> I know it's passing the crypto check, because I can flash a modified fastboot bootloader in the same way and it works still.

17:32 <digetx> pgwipeout[m]: I don't know, but you may try to flash something simple as reboot via PMC one-line assembly, it either should work or there is indeed some extra check

21:38 <DavidHeidelberg> digetx are u ok with changing "samsung_p3,isa1200_vibrator" to "samsung,isa1200_vibrator" ?

21:38 <DavidHeidelberg> since first string should be vendor, samsung_p3 doesn't seems to identify anything (also tried search it on the web)

21:40 <digetx> shouldn't it be imagis,isa1200?

21:42 <DavidHeidelberg> oh, yes. Right, are u ok with that change?

21:42 <DavidHeidelberg> to imagis

21:43 <digetx> yes

21:46 <m4t> pgwipeout[m]: how are you flashing the encrypted vs unencrypted bootloader?

21:48 <m4t> same way eh

21:48 <m4t> are you trying to pre-encrypt it with openssl?

21:53 <pgwipeout[m]> m4t: Encrypted with the nvencrypt tool for the n7 off device unbrick. Allows use of nvflash as is.

21:54 <pgwipeout[m]> Also tried using fastboot to flash it (the cracked bootloader lets flashing unsigned images).

21:55 <pgwipeout[m]> That was the original way I tried, but it would fail to boot, I thought it was a problem with the encryption.

21:55 <pgwipeout[m]> But then I found that encrypted and sending it via nvflash it doesn't even fire off, but sending it unencrypted with security disabled it does.

21:56 <pgwipeout[m]> Examining the official unencrypted fastboot bootloader I see a few extra things in there. One is TZ, but I think the other is the signed and encrypted miniloader.

21:57 <pgwipeout[m]> There's not much left of the L4T documentation from the T3 days but it seems to suggest there needs to be a miniloader signed by nvidia included in the encrypted bootloader. Unencrypted nvflash sends this miniloader during the initialization sequence.

21:57 <pgwipeout[m]> But the L4T links from that era are broken.

21:58 <digetx> webarchive doesn't help?

21:59 <pgwipeout[m]> digetx: No, these were behind the nvidia developer login.

21:59 <pgwipeout[m]> I know it's flashing correctly because I successfully encrypted and flashed the cracked boot loader. The BCT is also getting updated correctly.

22:03 <m4t> iirc n7 doesn't expect bootloader to be encrypted, ouya does

22:03 <m4t> not sure if it's a fuse, odmdata, or some flag in the partition

22:14 <m4t> i don't think it's just a header prepending the bootloader data though

22:16 <m4t> ouya has odmdata 0x40099000 and n7 is 0x40000000, i can't find any docs on it though. and messing with it is a good way to brick the board ime.

22:25 <pgwipeout[m]> Oh no, N7's bootloader is encrypted as well.

22:26 <DavidHeidelberg> digetx https://github.com/grate-driver/linux/pull/76 and good night :)

22:29 <DavidHeidelberg> best score 2864 lines of output. so ~= 2500 warnings remains

22:31 <m4t> pgwipeout[m]: oh, well the only flash dumps i have show it unencrypted

22:32 <m4t> maybe they're just of failed attempts to reflash

22:35 <m4t> oh duh, looking at another more recent one, it's definitely encrypted :P

22:35 <m4t> this must've been from an early attempt to reflash using nvflash

22:39 <m4t> pgwipeout[m]: is there a reason you're trying to pre-encrypt? to avoid the intermediate step of a bootloader with signing disabled?

22:52 <m4t> cbootimage seems to use a hard-coded aes key of 0x0, i wonder if you could modify it to use the sbk

22:53 <pgwipeout[m]> cbootimg expects a ssl key pair, it doesn't like a text sbk.

22:53 <pgwipeout[m]> In the absence of a SBK, 0x0 is used.

22:54 <pgwipeout[m]> Because essentially encryption is always on, it's just a default key of 0x0.

22:55 <m4t> isn't rsa only for newer chips? cbootimage seems to support t20 t30

22:55 <pgwipeout[m]> <m4t> "pgwipeout: is there a reason you..." <- Originally I had an issue with flashing u-boot this way, I thought it might only work with the original bootloader. Now that I have working nvflash I realize that isn't the issue, and there's something that only breaks when u-boot is encrypted.

22:55 <pgwipeout[m]> T30 is an AES-128 key.

22:56 <pgwipeout[m]> Essentially, if you have working nvflash native, you know your encryption is good.

22:56 <m4t> where does cbootimage expect rsa for t30?

22:57 <m4t> it seems to just encrypt with aes and put the hash into the bct

22:59 <m4t> samples/sign.sh does but that's for t124/t210

23:01 <m4t> "encrypts" since its 0x0

23:01 <pgwipeout[m]> I'm confused where rsa is coming from.

23:02 <m4t> ssl key pair?

23:02 <m4t> https://github.com/NVIDIA/cbootimage/blob/master/src/crypto.c#L51-L53 maybe it's possible to just change that to sbk, or modify it to accept sbk from command line parameter

23:02 <m4t> it doesn't seem to require that, maybe i'm missing something

23:03 <pgwipeout[m]> Yeah, the nvidia tools expect either a sbk in a signed key file or a pre encrypted blob.

23:03 <m4t> oh you mean nvflash?

23:05 <m4t> were you able to generate a working blob using wheelie?

23:05 <pgwipeout[m]> Any tool that supports flashing encrypted bootloaders. Can't say if cboot supports it on T20 or T30, but nvflash does. TegraRCM does not.

23:06 <pgwipeout[m]> Wheelie doesn't generate blobs, it consumes them like nvflash does.

23:07 <pgwipeout[m]> It might also be the lack of ATF that's killing it. encryption may demand ATF.

23:07 <m4t> wheelie generates blobs

23:08 <pgwipeout[m]> Wheelie is just a reimplementation of the nvflash protocol.

23:08 <pgwipeout[m]> https://github.com/AndroidRoot/wheelie

23:08 <pgwipeout[m]> " Support pre-encrypted blobs"

23:08 <pgwipeout[m]> "Supports bootstrapping a pre-encrypted bootloader from USB BootROM (APX)"

23:09 <m4t> er, sorry i meant flatline

23:10 <m4t> *were you able to generate a working blob using flatline?

23:10 <m4t> i couldn't get it to work

23:10 <pgwipeout[m]> From a quick search flatline requires on device encryption working.

23:11 <pgwipeout[m]> nvcrypttools supports off device encryption.

23:15 <pgwipeout[m]> (Though getting it to build is a bloody nightmare)

23:15 <m4t> the blob flatline generates should allow nvflash to be used without any patched miniload etc.

23:15 <m4t> was wondering if you got that part to work

23:16 <pgwipeout[m]> Yes, but all of the on device solutions are unsupported, dead, or don't work.

23:16 <pgwipeout[m]> And yes, I've gotten nvflash native working with nvcrypttools off device generation.

23:26 <m4t> were you able to generate a blob for ouya?

23:26 <m4t> https://forum.xda-developers.com/t/tutorial-generate-blobs-for-a-bricked-nexus-7-without-another-n7-or-tegra30-device.4305955/

23:27 <pgwipeout[m]> Yes, that's nvcrypttools.

23:27 <m4t> ya i couldn't get it to compile before

23:28 <m4t> that was jevinskie's fork though i think

23:29 <m4t> as is this. it was issues with mbedtls iirc

23:35 <pgwipeout[m]> Yeah, that's what I ran into as well it's a problem with the configuration of the distro mbedtls. It builds on ubuntu 21.04. I'd like to fork it and make a few fixes, like the file generation leaves a lot to be desired.

23:36 <m4t> doesn't seem to like gcc-10

23:36 <m4t> i think i tried building on alpine then gave up

23:36 <m4t> ouya:~/devel$ ls -ld nvcrypttools/

23:36 <m4t> drwxr-xr-x 8 matt matt 3488 Mar 23 2020 nvcrypttools/

23:56 <m4t> oh i got it to compile on alpine actually, the gcc command line needed libraries placed after the .o